I'm pretty sure that their internal dataset is far more rich/detailed than the public-facing "Zestimate".

Traffic-data alone, segmenting the buyers clicking on each listing and their expressions of intent, would give them an incredibly-actionable trading/training signal.

Zillow might see, nationally, the nationwide (and regional) direction of buyer and seller sentiment 2-4 weeks ahead of all but the largest regional brokerages.

That's all everything ever is anymore.

You don't need to be a good stock picker if you know that someone sent a buy out, and you can get there first, buy it and then sell it to them.

You don't need to find good domain names, just wait for someone to search if one is available and then buy it.

Or when you Google for Verizon and the first thing you see is an ad for verizon.com. Most people will click the ad instead of the first listing.